Fig. 7: Influence of denoising on pixel-classification-based membrane segmentation in TEM data. | Nature Communications

Fig. 7: Influence of denoising on pixel-classification-based membrane segmentation in TEM data.

From: An interactive ImageJ plugin for semi-automated image denoising in electron microscopy

Fig. 7

The first column shows the input, the top image additionally shows a fraction of the labels that were used for training. The second column shows the probability output map of the random forest pixel classifier from ilastik. The third column shows the segmentation result by thresholding the probability maps, and the fourth column the ground truth segmentation (the Dice coefficient is shown in the upper left corner). Noise artifacts are clearly visible in the segmentation and can be avoided by denoising as a pre-processing step. Notice that denoising can even improve the segmentation result.

Back to article page